Additional file 1:

Animation to Demonstrate Membership Propagation. This animation shows how the influence of memberships of CSOs and outliers propagate through a network formed by neighborhood relationships, and how a gene gets a fuzzy membership by the balanced influence from its neighbors.

Format: AVI Size: 1.2MB Download file

Playing the movie within this page requires QuickTime 3 or later and JavaScript. Read more

Additional file 2:

Additional Note. This note includes a proof for the heuristic optimization procedure used in FLAME and a rough time complexity analysis of the FLAME algorithm.

Format: PDF Size: 60KB Download file

This file can be viewed with: Adobe Acrobat Reader

Additional file 3:

Empirical time complexity comparison of FLAME with other algorithms. This comparison is done on the hypoxia dataset with 57 samples. Gene subsets of different sizes are obtained by choosing genes with the highest variations.

Format: PNG Size: 54KB Download file

Additional file 4:

Clustering validation and comparison by range FOM. a, range FOM on the reduced peripheral blood monocyte dataset. b, range FOM on the reduced hypoxia response dataset. c, range FOM on the reduced yeast cell cycle dataset.

Format: PNG Size: 252KB Download file

Additional file 5:

Annotation matrices. Annotation matrices of 44 clusters (rows) across 230 GO terms (columns) obtained from the mouse tissue dataset. The color scale indicates the number of counts for each GO term (column) in each cluster (row). Matrices obtained by FLAME, hierarchical, k-means, fuzzy SOM and fuzzy C-means clustering mouse tissue dataset are shown, as indicated. The grey color indicates zero counts for a given GO term in a given cluster. The average annotation profile can be detected as a row without grey cells and with name "dataset" in the right part.

Format: PNG Size: 282KB Download file